Each lesson runs the same tiny LLM neuron, y = ReLU(0.5·x₁ − 0.25·x₂ + 0.1) with x = [1.5, 2.0], through a real toy version of one scheme, step by step and animated. Compare how every scheme handles the linear part and the ReLU.
Paper database → every paper in the archive with metadata, extracted details and full-text keyword counts (SQLite + MkDocs)